Radar‐based cell tracking with fuzzy logic approach

Abstract

ABSTRACTThis study presents a new algorithm Fuzzy logic Algorithm for Storm Tracking (FAST) for tracking convective cells in 3D Cartesian co‐ordinates. A fuzzy logic approach was applied to track cells with objectively determined membership functions (MFs) and weights that were derived from a large set of convective cell tracks (2326 pairs of convective cells). Three feature parameters (cell motion speed, area change ratio and axis transformation ratio) were selected by Bayesian information criterion and combined through fuzzy logic. FAST was evaluated with three skill scores using 2561 (2456) pairs of convective cells with (without) mergers and splits. The performance was tested in comparison with Thunderstorm Identification, Tracking, Analysis, and Nowcasting (TITAN) as a baseline method. For reference tracks without merger and split, the critical success index (CSI) of FAST reaches about 0.89. FAST shows better skill scores (CSI = 0.92 and 0.89 at speed limitations of 16.1 m s−1 and 25.0 m s−1, respectively) than those of TITAN (CSI = 0.89 and 0.82 at the same speed limitations) and is independent of the number of convective cells and the speed limitation. When mergers and splits are considered in the reference, the tracking performance of FAST (FASTMS) with a merger and split process is always better than that of TITAN (TITANMS) with a merger and split process and is independent of speed limitations. When FAST (=FASTnoMS) does not include a merger and split process, its tracking performance is better than that of TITANMS at speed limitation of > 20 m s−1.